Diabetes clusters and risk of diabetes-associated diseases – Authors' reply

We thank Shufang Liu and Wenquan Niu for their interest in our Article,1Zaharia OP Strassburger K Strom A et al.Risk of diabetes-associated diseases in subgroups of patients with recent-onset diabetes: a 5-year follow-up study.Lancet Diabetes Endocrinol. 2019; 7: 684-694Summary Full Text Full Text PDF PubMed Scopus (235) Google Scholar which applied the clustering algorithm proposed by Ahlqvist and colleagues2Ahlqvist E Storm P Käräjämäki A et al.Novel subgroups of adult-onset diabetes and their association with outcomes: a data-driven cluster analysis of six variables.Lancet Diabetes Endocrinol. 2018; 6: 361-369Summary Full Text Full Text PDF PubMed Scopus (1009) Google Scholar to patients with diabetes in the German Diabetes Study (GDS).3Szendroedi J Saxena A Weber KS et al.Cohort profile: the German Diabetes Study (GDS).Cardiovasc Diabetol. 2016; 15: 59Crossref PubMed Scopus (82) Google Scholar We would like to respond to their comments on the methodological aspects of our study. First, the GDS provided the unique opportunity to assess all necessary variables in patients both during the first year upon diagnosis of diabetes and 5 years later. This approach made it possible to monitor changes in cluster allocation and detect lower reproducibility in the severe insulin-deficient diabetes (SIDD) cluster. Notably, the observational design of the GDS does not control for changes in lifestyle and medications over this period. As we reported,1Zaharia OP Strassburger K Strom A et al.Risk of diabetes-associated diseases in subgroups of patients with recent-onset diabetes: a 5-year follow-up study.Lancet Diabetes Endocrinol. 2019; 7: 684-694Summary Full Text Full Text PDF PubMed Scopus (235) Google Scholar the switching from one cluster to another is likely to be due to improved glycaemic control, possibly resulting from intensified treatment. We would therefore explain this finding by differences in the applicability of this algorithm to specific patient clusters and by changes in metabolic and other variables over time rather than by differences in the suitability of this clustering algorithm for different cohorts. Nevertheless, cohort-specific variances might affect the reproducibility of the proposed clustering2Ahlqvist E Storm P Käräjämäki A et al.Novel subgroups of adult-onset diabetes and their association with outcomes: a data-driven cluster analysis of six variables.Lancet Diabetes Endocrinol. 2018; 6: 361-369Summary Full Text Full Text PDF PubMed Scopus (1009) Google Scholar due to the inclusion and exclusion criteria of the GDS.3Szendroedi J Saxena A Weber KS et al.Cohort profile: the German Diabetes Study (GDS).Cardiovasc Diabetol. 2016; 15: 59Crossref PubMed Scopus (82) Google Scholar As we did not use de novo clustering in our GDS cohort, but applied the already established algorithm,2Ahlqvist E Storm P Käräjämäki A et al.Novel subgroups of adult-onset diabetes and their association with outcomes: a data-driven cluster analysis of six variables.Lancet Diabetes Endocrinol. 2018; 6: 361-369Summary Full Text Full Text PDF PubMed Scopus (1009) Google Scholar the suggested evaluation of the effectiveness of the clustering algorithm was not feasible in this context.4Hennig C Cluster-wise assessment of cluster stability.Comput Stat Data Anal. 2007; 52: 258-271Crossref Scopus (378) Google Scholar Nevertheless, applying a numerical estimate for the stability of cluster membership between baseline and follow-up, in addition to the Sankey diagram presented in our original publication,1Zaharia OP Strassburger K Strom A et al.Risk of diabetes-associated diseases in subgroups of patients with recent-onset diabetes: a 5-year follow-up study.Lancet Diabetes Endocrinol. 2019; 7: 684-694Summary Full Text Full Text PDF PubMed Scopus (235) Google Scholar yielded a value of 0·62 (95% CI 0·58–0·68) for the Cramer's V coefficient,5Agresti A Kateri M Categorical data analysis.in: Lovric M International encyclopedia of statistical science. Springer Berlin Heidelberg, Berlin2011: 206-208Crossref Google Scholar indicating overall good stability. Second, we would like to clarify that, by design, several patients were and are still not eligible for the follow-up examination, simply because they have a known diabetes duration of less than 5 years within the ongoing GDS.3Szendroedi J Saxena A Weber KS et al.Cohort profile: the German Diabetes Study (GDS).Cardiovasc Diabetol. 2016; 15: 59Crossref PubMed Scopus (82) Google Scholar 464 (42%) of 1105 patients had a disease duration of less than 5 years. Moreover, only 76 (7%) patients did not consent to all tests and were therefore excluded from the follow-up analysis because of incomplete data. Notably, the follow-up study population was representative of the total cohort at baseline, with regard to the main clinical and metabolic features. MR reports personal fees from Eli Lilly, Novo Nordisk, Poxel SA, Boehringer-Ingelheim Pharma, Terra Firma, Sanofi US, Servier Labatories, Prosciento, and Fishawack Group. OPZ, OK, KS, VB, and JS declare no competing interests. Diabetes clusters and risk of diabetes-associated diseasesWe read with interest the study by Zaharia and colleagues,1 who adopted the diabetes five-cluster algorithm proposed in 2018 by Ahlqvist and colleagues2 and characterised a cohort of patients with different degrees of whole-body and adipose-tissue insulin resistance. The authors1 reported distinct metabolic alterations and specific risk patterns for the development of diabetes-related comorbidities and complications after 5 years of follow-up. Here, we comment on two methodological aspects of this study. Full-Text PDF Risk of diabetes-associated diseases in subgroups of patients with recent-onset diabetes: a 5-year follow-up studyCluster analysis can characterise cohorts with different degrees of whole-body and adipose-tissue insulin resistance. Specific diabetes clusters show different prevalence of diabetes complications at early stages of non-alcoholic fatty liver disease and diabetic neuropathy. These findings could help improve targeted prevention and treatment and enable precision medicine for diabetes and its comorbidities. Full-Text PDF